Re: Request for Ada Coding Standards

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Mike Brenner says

<<Why ask people to standardize syntax instead of the things that
make the code work? Why not ask people to:

   document the preconditions and boundaries of the software
   not delete customer service requests until released
   minimize coupling (such as pointers or references to global variables)
   maximize cohesion (each unit does one thing in a simple manner)

but do it any style they like. They will produce higher quality at
a lower life-cycle cost if you measure the above 4 items instead of
lines of code, capitalization, indentation, number of lines per unit,
bulky unit headers, meaningless comments, complexity metrics that
penalize nested CASE statements, or complexity metrics that give
equal penalty to multiple entrances (BAD) and multiple exits (GOOD),
in my opinion. I am not aware of any research that shows a reduction
of life-cycle cost for measuring syntax compliance over my 4 suggested
standards above. As a matter of fact, I am not aware of any research
that shows a reduction in life-cycle cost for having a syntax standard
over any ONE of my 4 suggested standards above.>>

Actually there *is* research that shows the benefit of uniform style,
at least in a COBOL environment, and it seems pretty clear that these
benefits are language independent.

Yes, we know that Michael does not think that coding standards are
worth while, but this is really such a far out view that I don't think
it is worth discussing (especially since it is unlikely to add anything
over the last time we did). The fact is that the great majority of the
programming community agrees that uniform coding standards are necessary,
so it seems a good idea to concentrate on WHAT standards should be used,
which was indeed the original posters interest, rather than warming over
previous discussions with Michael on WHETHER standards are uself. The
latter discussion is unlikely to add anything new.

On the other hand, I cannot imagine any research which shows that
"not deleting customer service requests until released" helps
code quality, or that "maximizing cohesion" helps code quality.
These ideas, while reasonable in some environments, are far too
vague.

Also I am not even sure that i agree with "each unit does one thing".
The idea of a package of related operations seems natural and desirable
to me, rather than splitting eafch function into a separate package.
\x1adp